Next-Generation CAR T-cell Therapies

Chimeric antigen receptors (CAR) are synthetic receptors that redirect T cells to target cancer cells in a major histocompatibility complex–independent manner. T cells engineered to express a CAR recognizing CD19 have led to unparalleled responses in a subset of otherwise refractory patients with B-cell malignancies (1). However, despite the tremendous impact of CD19-directed CAR T-cell therapy, primary resistance or relapse is frequently observed with varying patterns and mechanisms of resistance across different entities.

Although the initial response rate is high, only approximately 50% of patients with pediatric and young adult acute lymphoblastic leukemia (ALL) treated with CAR T cells will be alive and disease-free 1 year after treatment. For patients with higher disease burden at the time of treatment, the event-free survival at 12 months drops to 31% (2). The assessment of antigen status reveals CD19-negative or CD19-low disease in a high proportion of relapses (2, 3). This pattern of resistance could be principally addressed by sequential administration of CAR T cells targeting a different antigen. For instance, patients with ALL who were treated with anti-CD22 CAR T cells after relapse with CD19-directed therapy achieved a 73% complete response rate, but two thirds of these patients later relapsed with reduced CD22 surface density (4). Current approaches designed to block the escape of resistant tumor clones include upfront infusion of multispecific CAR T-cell products targeting several antigens simultaneously, achieved by either pooling of different CAR vectors, a single vector encoding two different CARs (bi-cistronic/dual CAR), or a single CAR construct (tandem CAR) incorporating two different single-chain variable fragments (scFv; Fig. 1A). Early clinical studies demonstrated feasibility of such approaches, for example, combining CD19/CD20 (5) or CD19/CD22 (6). However, such approaches come with the challenge to construct optimal CARs that retain efficacy for all targets. For example, Spiegel and colleagues lost potency against CD22 in their tandem construct, indicated by undiminished CD22 expression in the relapse situation compared with baseline (6), which was not predicted in preclinical models (7) and contrasts with the observed immune pressure on CD22 with a monospecific construct (4). A potential strategy to address antigen-dim relapse relies on fine-tuning the antigen density threshold for CAR-mediated T-cell activation. This can be achieved by adapting the scFv affinity of the CAR in accordance with a certain antigen expression level, whereby a high-affinity CAR can principally detect a dim antigen (8), or by changing/altering domains of the CAR such as the transmembrane/hinge region or the costimulatory domain, for example, via manipulation of the number of immunoreceptor tyrosine-based activation motifs (ITAM; refs. 9–11). However, this threshold needs to be exquisitely adapted to the respective clinical scenario and carefully studied, as CAR affinity modulation will potentially impact T-cell effector function, kinetics of immunotoxicity, and the threshold for on-target, off-tumor toxicity for less specific antigens than CD19. More so, the superiority of a lower-affinity CD19 CAR above the clinically established FMC63 binder in an antigen-abundant scenario has been suggested (12).

Antigen-dim relapse has also been observed in patients with multiple myeloma treated with anti–B-cell maturation antigen (BCMA) CAR T cells (13). Pharmacologic modulation of target antigen expression levels is a promising strategy to combat antigen-dim tumor relapse. For instance, small-molecule γ-secretase inhibitors can increase BCMA surface levels on myeloma cells (14) by blocking its cleavage and are currently under clinical investigation together with BCMA-directed CAR T cells (e.g., NCT03502577).

Unlike therapies with cytotoxic agents and small molecules, cell therapies are generally agnostic to tumor genotype, given expression of the selected CAR target antigen. However, within B-cell tumors, despite similar target density of CD19, there is tumor-type susceptibility to CD19-directed CAR T cells. Patients with large B-cell lymphoma treated with anti-CD19 CAR T cells show lower complete response (CR) rates—between 40% and 50%—as compared with patients with B-cell ALL—between 70% and 90%. Responses in patients with lymphoma tend to be durable; however, approximately 30% of patients who relapse demonstrate antigen downregulation (15). Reasons for failure to achieve initial response are complex and encompass associations with clinical factors such as high tumor burden (16) and T-cell–intrinsic deficits as well as features of the tumor microenvironment (TME) such as the accumulation of myeloid-derived suppressor cells and interferon signaling (17).

Although antigen escape is a major resistance mechanism in pre–B-cell ALL, relapse is rare in the more mature B-cell–derived malignancy chronic lymphocytic leukemia (CLL). On the contrary, a high rate of initial resistance, linked to T-cell–intrinsic deficits in the patient, is observed. Durable responses can still be achieved in approximately 30% of patients with CLL (18). CD8⁺ T cells of patients with CLL demonstrate impaired mitochondrial fitness and reduced glucose uptake upon stimulation (19). Patients with CLL who did achieve a complete remission were found to have enhanced mitochondrial biogenesis in correlation with better proliferation and persistence (19). Sustained remissions in patients with CLL are furthermore associated with a higher frequency of memory-like T cells (CD27⁺CD45RO⁻CD8⁺) and STAT3-related gene signatures in the patient apheresis product (20).

Whenever a lack of T-cell fitness is presumably the reason for CAR T-cell treatment failure, two main strategies have been explored as possible solutions: It has been demonstrated for patients with multiple myeloma that T cells collected early during the disease course with less exposure to cytotoxic treatment retain better fitness and proliferative capacity compared with those collected late from relapsed or refractory patients (21). Strategies to leverage these insights will incorporate manufacturing of autologous CAR T-cell products from less dysfunctional T cells either by obtaining apheresis products early during disease course or by shifting CAR T-cell therapy into an earlier line of treatment. The first examples of this strategy have already demonstrated an enhanced proliferative potential of CAR T-cell products alongside promising response rates (e.g., ZUMA-12 and NCT03761056) in patients with lymphoma (22). Alternatively, a potential source of healthy T cells is the use of allogenic donors with additional potential benefits as discussed below (23).

Another elegant way to improve CAR T-cell function is to confer the ability to secrete active cytokines for beneficial immunomodulation, so-called T cells redirected for universal cytokine killing (TRUCK; Fig. 1B). This way, robust T-cell activation can become less dependent on an external “signal 3.” Examples are IL12 and IL18 and improved antitumor effects of these engineered cytokines and receptors as demonstrated in xenograft and syngeneic mouse models (24, 25). IL18 not only benefits CAR T cells in an autocrine fashion but also modulates host immune cells via paracrine effects (26). These results have motivated several ongoing clinical trials. The recent characterization of IL18 binding protein (IL18BP) in the TME as a “secreted negative immune checkpoint” highlights the potential for further refinement of this approach (27). In order to escape IL18BP binding, CAR T cells could be either engineered to secrete a decoy-resistant IL18 (DR-18; ref. 27) or designed to activate IL18 signaling independently of secreted IL18 by utilizing a GM-CSF/IL18 switch receptor (28).

Broadening the scope of CAR T-cell therapies beyond CD19 (or BCMA)-positive hematologic malignancies seems within reach, as there are several other potential surface antigens within the cluster of differentiation system with manageable potential for on-target, off-tumor toxicity largely confined to hematopoietic cells, such as CD5, CD7, and CD30. For targets for which ablation of lineage antigen–bearing cell populations is not clinically tolerated, a tandem therapy combining targeted CAR T cells with a hematopoietic stem cell transplant genetically engineered to lack the target antigen has been proposed. For example, CD33 is highly expressed on most acute myeloid leukemia blasts but is also expressed on myeloid precursor cells, foreshadowing unacceptable toxicity in the form of persistent myelosuppression if targeted with CAR T cells (29). Interestingly, CD33 is dispensable for myeloid function, and by means of subtraction, a leukemia-specific antigen can therefore be created by combining CD33-directed CAR T cells with a CD33-deleted hematopoietic stem cell transplantation (HSCT; refs. 30, 31; Fig. 1C).

Similarly, targeting T-cell malignancies with CAR T cells would threaten the healthy T-cell compartment as a consequence of shared target gene expression. Even more prohibiting, CAR T-cell expansion is prone to fail during manufacturing due to antigen-mediated self-killing or fratricide. However, successful production of fratricide-resistant CAR T cells targeting CD7 or CD3 and CD7 can be accomplished by knocking out these targets before in vitro expansion (32, 33). An interesting byproduct of this strategy is that it also leads to self-enrichment for the desired edits in the final CAR T-cell product by residual fratricide against unedited cells (33). As a limitation, such a treatment strategy would likely need to incorporate a subsequent stem cell transplantation in remission to restore the T-cell compartment. Also, the therapy would likely rely on an allogeneic CAR T-cell product to prevent contamination by malignant T cells.

Universal CAR T cells address a critical barrier in the field; healthy allogenic donor T cells with superior fitness have the potential to improve clinical responses (Fig. 2). “Off-the-shelf” allogeneic CAR T-cell products also provide an opportunity to streamline manufacturing and allow treatment of rapidly progressing patients who are unable to bridge several weeks of manufacturing. By reducing costs, universal CAR T cells will potentially broaden patient access to this treatment modality. The major hurdles associated with allogeneic CAR T-cell products are the risk for graft-versus-host disease (GvHD) mediated by the endogenous T-cell receptor (TCR) and the risk of T-cell graft rejection by the host immune system, resulting in limited persistence (23). A potential solution to tackle these hurdles is offered by advancements in genome editing, mostly based on CRISPR–Cas nucleases, and its technological successors (34). These technologies are also suitable to address other challenges of the field, such as the modulation of T-cell exhaustion (35). Genome editing can be utilized in allogeneic CAR T cells to remove the endogenous TCR, eliminating the risk of GvHD, and beta-2-microglobulin (B2M) to ablate HLA class I molecule expression for preventing graft rejection mediated by host T cells. The feasibility of this concept has already been demonstrated in the two recently published UCART19 trials (36), but CAR T-cell persistence was limited and retained only in patients with severe immunosuppression mediated by the anti-CD52 antibody alemtuzumab.

Clinical applications involving genome editing come with the major concern for off-target editing, which is the unintended and uncontrolled modification of DNA loci beyond the intended edit that might alter cellular function (34). This risk is inherent to the mechanism of action by which CRISPR–Cas nucleases induce DNA double-strand breaks (DSB), which poses a risk for translocations proportional to the number of targets in a multiplex setting, leading to theoretical concerns of malignant transformation (37). Also, DSBs likely imply negative effects on T-cell proliferation and fitness (38, 39). Recent techniques, such as prime and base editing, enable genome editing without the creation of DSBs, and their principal applicability for T cells has been demonstrated (40, 41). Extensive work has been conducted to understand most of those off-target effects and to make them predictable and/or measurable (34) in order to mitigate their risks. In CAR T-cell applications, the components of the editing machinery are usually only transiently exposed and applied ex vivo, contributing to safety. Further, in a pilot study, CRISPR–Cas9-modified T cells containing three targeted loci on three different autosomal chromosomes were infused into patients with relapsed/refractory cancer, and modified cells persisted for 9 months without clinical toxicities or safety concerns (42).

In addition to overcoming host T-cell–mediated rejection of allogeneic cell products, it is likely that infused cells will require additional genome engineering to avoid rejection by the innate immune system. For example, in mice, host natural killer (NK) cells and macrophages rapidly destroy infused allogeneic T cells via distinct mechanisms (43–45). A major remaining challenge facing the field is to further elucidate the mechanisms of allogeneic T-cell rejection in the human immune system.

Beyond T cells, NK cells offer a promising alternative approach for an allogenic cellular immunotherapy. NK cells are part of the innate lymphoid cell lineage and play a critical role in immune defense against tumors and pathogens (46). They can be adoptively transferred into patients as an allogeneic product without the need for gene editing to avoid the high risk of GvHD. Further, if transduced with a CAR directed at a tumor-specific antigen, NK cells can attack tumor cells via antigen-dependent cytotoxicity, as well as through their activating NK receptors, and thus are an attractive complement to allogeneic CAR T-cell therapy. NK cells are generally less abundant in peripheral blood compared with T cells, posing challenges to clinical-scale manufacturing. Cord blood, induced pluripotent stem cells, embryonic stem cells, NK-92 cells, and an NK cell line have been used as alternate sources of NK cells for adoptive cell transfer (ACT; ref. 47). In addition, the proliferative capacity and persistence of NK cells are comparatively less than those of T cells; however, treatment with cytokines such as IL2, IL15, and IL21 can extend the NK life span. In addition, the risk of toxicities like cytokine release syndrome or neurotoxicity may be reduced in comparison with CAR T-cell therapies, as activated NK cells secrete low amounts of IL6 or IL1β (46). Currently, approximately 10 clinical trials in ClinicalTrials.gov are recruiting patients for the treatment of cancers, including various B-cell malignancies, multiple myeloma, acute myeloid leukemia, and solid tumors with CAR NK-cell products. Notably, in a recent phase I/II clinical trial, patients with refractory non-Hodgkin lymphoma or CLL were administered CD19 CAR NK cells armored with IL15 (48). Patients (73%) experienced objective responses; however, the durability of response could not be assessed because other therapies were delivered as early as 30 days after the infusion of CAR NK cells (48). Patients experienced high-grade, transient myelotoxicity, which may have been caused by the lymphodepleting chemotherapy, but not cytokine release syndrome, neurotoxicity, or GvHD. The product was manufactured from HLA-mismatched and when possible killer immunoglobulin-like receptor ligand–mismatched cord blood and was freshly infused after manufacturing (48). In the future, advancement of clinical-grade CAR NK products that retain efficacy and persistence after cryopreservation and thaw will realize “off-the-shelf” products for CAR NK therapy, increasing treatment options for patients (49).

Currently, the efficacy of CAR T-cell therapies targeting solid tumors is restrained by the quantity, persistence, and functionality of transferred T cells and by the heterogeneity and immunosuppressive nature of the TME (50–69). Addressing these challenges is critical to future successes in ACT. Dose-limiting toxicities associated with CAR T-cell recognition of the low level of target expression on normal tissue (“on-target, off-tumor”), and toxicities associated with strong CAR T-cell efficacy, pose barriers that limit therapeutic dosing of CAR T cells. For example, several ongoing clinical trials designed to treat MSLN-expressing tumors have highlighted an association between antitumor efficacy and lung-associated dose-limiting toxicity. Interim results from the phase I portion of a clinical trial administering anti-MSLN–directed TCR T cells, gavo-cel (NCT03907852), for the treatment of MSLN-expressing solid tumors reveal that, following lymphodepletion, patients infused with gavo-cel achieved a 25% objective response rate; however, dose-limiting pulmonary toxicities were reported. Likewise, in a phase I dose-escalation trial targeting MSLN-expressing refractory malignant cancers, a patient with mesothelioma experienced grade 5 respiratory failure after infusion of fully human MSLN-directed CAR T cells at a dose of 1× 10⁸ to 3 × 10⁸ cells/m² (NCT03054298). In both studies, these grade 5 toxicities prompted a dose de-escalation, and no serious events were reported at the lower dose. In patients with malignant pleural disease, this lung-associated toxicity was bypassed by the regional delivery of anti-MSLN CAR T cells (0.3 × 10⁶ to 60 × 10⁶ cells/kg) followed by PD-1 blockade, resulting in stable disease for ≥ 6 months in eight of 18 patients, with two patients exhibiting a complete metabolic response on PET scan (NCT02414269; ref. 70). The regional delivery of CAR T cells targeting MSLN-expressing cancers is also under evaluation at other centers (NCT03054298 and NCT03323944).

Further, when the efficacy of CAR T-cell therapies targeting solid tumors reaches a therapeutic level, patients can experience severe toxicities. In particular, significant declines in prostate-specific antigen levels have been observed in patients with castration-resistant metastatic prostate cancer treated with prostate-specific membrane antigen (PSMA)–directed CAR T cells (NCT04249947) or TGFβ-insensitive PSMA-targeted CAR T cells (NCT03089203). In both studies, increases in CAR T-cell doses led to significant antitumor activity, accompanied by patient mortality. The causes of the patient deaths are currently unclear, but may be associated with macrophage activation syndrome (MAS) or immune effector cell-associated neurotoxicity (ICANS; refs. 71, 72), although on-target, off-tumor toxicity is theoretically possible. It is conceivable that new tocilizumab-like drugs will be needed to mitigate ICANS and MAS toxicity associated with robust CAR T-cell therapy to allow patients with solid tumors to tolerate an effective dose (73).

Another approach to bypassing on-target, off-tumor–related toxicities caused by aberrant CAR T-cell destruction of normal tissues is the use of dual antigen–sensing strategies that permit more precise immune recognition of tumors. One such example is the SynNotch receptor (Fig. 1C). This synthetic circuit is constructed to allow T cells to integrate combinatorial antigen expression data to discriminate normal and tumors cells. In the AND-gated version of a two-receptor circuit, activation of one receptor incudes the transcription of a second receptor (CAR or TCR), killing tumor cells that simultaneously express both antigens and sparing normal tissue that expresses only one antigen (74, 75). In contrast, NOT-gated SynNotch circuits are designed to minimize CAR T-cell recognition of normal cells by designing a circuit that combines a CAR recognizing the shared target antigen and an inhibitory CAR-preventing killing when the second antigen, present only on normal bystander cells and not tumor, is present. Careful selection of antigen pairs specific to the CAR circuit and target disease permits the construction of CAR T cells with limited on-target, off-tumor toxicity in vitro and in vivo mouse models. The recent development of fully humanized circuits, termed “synthetic intramembrane proteolysis receptors (SNIPR),” that mirror the function of first-generation SynNotch receptors supports the clinical translation of CAR T cells with the capacity to sense multiple combinatorial antigens on a cancer cell by reducing potential immunogenicity. However, this approach does not abrogate the risk of tumor relapse due to antigen escape, including antigen loss or downregulation.

Unlike CAR T-cell therapy directed against CD19, adoptively transferred CAR T cells directed against solid tumors undergo limited homeostatic expansion in the blood and encounter tumor antigen in an immunosuppressive TME where the number of CAR T cells may be insufficient to eradicate disease (62). Local administration of CAR T cells is being investigated to address the limited number of CAR T cells reaching and expanding in the tumor (67, 70, 76, 77). Additionally, in vivo antigen-specific CAR T-cell expansion may enable dosing at sufficient levels to promote antitumor responses. One promising synthetic approach to selectively expand CAR T cells and omit systemic toxicity of native IL2 administration is the engineering of IL2 cytokine receptor orthogonal (ortho) pairs, which interact with each other but not their native cytokine receptor pairs (refs. 27, 78; Fig. 1B). Nanoparticle RNA and peptide vaccines offer an alternate approach to enhance the in vivo expansion of CAR T cells targeting solid tumors. Both CAR antigen-encoding RNA-lipoplexes (RNA-LPX), and CAR T ligands bound to albumin-binding phospholipid polymers (amph-ligand) vaccines deliver CAR antigen to lymph nodes, decorating the surfaces of antigen-presenting cells with the CAR ligand (79, 80). Currently, a phase I/IIa dose-escalation trial is open to evaluate the safety and preliminary efficacy of CLDN6 CAR T ± CLDN6 RNA-LPX in patients with CLDN6-positive relapsed or refractory advanced solid tumors (NCT04503278).

When CAR T cells enter an immunosuppressive TME, they acquire a dysfunctional or exhausted T-cell state that limits the therapeutic efficacy of CAR T cells targeting solid cancers (81). One mechanism driving the dysfunction of CAR T cells in the solid TME is plasticity of the CD8 T-cell state, where continuous antigen exposure promotes a CD8⁺ T-to-NK–like T-cell transition (82). The identification and CRISPR–Cas9-mediated knockout of genes that restrain tumor immunity have enabled the production of next-generation CAR T cells (42, 82–100). However, the extent to which the knockout of individual genes can prevent T-cell dysfunction is unclear. Recent studies suggest that T cells undergo a transition from a plastic dysfunctional state in which therapeutic reprogrammability is possible to a fixed dysfunction state that is resistant to reprograming (101). In particular, T cells exposed to chronic antigen stimulation are not able to fully recover after tumor clearance and acquire epigenetic changes or scars that can limit future immune responses (102, 103). These data highlight the potential of therapies promoting epigenetic remodeling to permit sustained CAR T-cell functionality. The possible therapeutic utility of epigenetic remodeling is further illustrated by mechanistic studies from a single patient with CLL treated with CAR T cells targeting CD19 (104). Lentiviral integration of the CAR transgene disrupted the patient's methylcytosine dioxy­genase TET2 gene, resulting in enhanced potency. Another strategy aimed at maintaining CAR T-cell functionality and limiting dysfunction-associated epigenetic remodeling is to modulate CAR expression or antigen exposure (105, 106). Unlike CAR T cells exhibiting constitutive CAR expression, regulated expression of the CAR by mechanisms such as transient rest, cessation of receptor signaling, or preventing CAR-mediated tonic signaling through SynNotch-controlled expression (107, 108) allows CAR T cells to avert exhaustion and maintain antitumor activity. In summary, future strategies aimed at mitigating the immunosuppressive effects of the TME will need to be multifactorial to target the complexity of cancer.

Engineered viruses, which specifically replicate in tumor cells without infecting healthy cells, have emerged as effective oncolytic agents in recent years. Although the molecular and cellular mechanisms are not fully understood, oncolytic viruses as a monotherapy are known not only to remodel the immunosuppressive TME through, for example, inflammatory effects at the injection site but also to induce systemic innate and adaptive antitumor immunity (109). These viruses can, however, also be harnessed for combinational approaches in cancer immunotherapy due to their ability to specifically transduce cancer cells with immunomodulatory or tumor-suppressor genes (Fig. 1D). In 2015, the first therapy of this kind, talimogene laherparepvec (T-VEC), a herpes simplex virus type 1–derived oncolytic agent, was approved (110). Response rates were further enhanced when T-VEC was combined with immune-checkpoint blockade (ICB) antibodies, and increased side effects were not observed (111). Mechanistically, oncolytic viruses can improve clinical responses to ICB by influencing the presence of tumor-infiltrating lymphocytes within the TME as well as the tumor mutational burden, which are known predictors of response to ICB (112, 113). Clinical trials testing the safety and feasibility of CAR T cells combined with oncolytic viruses are currently ongoing (NCT03740256 and NCT05057715).

A novel approach to modify the TME is the use of CAR-engineered macrophages (Fig. 1D), which have demonstrated efficacy in animal models (114). Macrophages are part of the innate immune system and actively participate in the immune response through phagocytosis and clearance of cellular debris (115). In addition, they are known to be potent antigen-presenting cells, inducing innate and adaptive immune responses. Based on these promising preclinical data, a first-in-human clinical trial on CAR macrophages targeting HER2-expressing tumors was recently started (NCT04660929), and several others have been announced. A limitation on the use of CAR macrophages is their terminally differentiated state, as, unlike CAR T cells, they do not proliferate after encountering their tumor target. Although macrophages are difficult to transduce with clinically proven lentiviral or retroviral vectors, CAR macrophages can be efficiently produced with clinically available Ad535 vectors, which are being used in the ongoing trials.

Combining stimulator of interferon genes (STING) agonists with CAR T cells could be another strategy to modify the TME and to activate innate/adaptive immunity, thereby improving overall CAR T-cell response. STING agonists trigger the expression of type I interferons and inflammatory genes, which leads to the activation of the innate immune defense and to the induction of the adaptive immune response. In a recently published preclinical study, the STING agonist DMXAA promoted CAR T-cell trafficking and persistence in a solid tumor model through alterations in the balance of innate and adaptive immunity as well as cytokine milieu within the TME (116). In another study, CAR T cells producing the RNA agonist RN7SL1 not only enhanced CAR T-cell function and expansion but also restricted the development of immune-suppressive cells and activated antitumor endogenous immunity. This resulted in control of different tumors, including tumors that had lost the CAR-targeted antigen (117).

Advances in next-generation CAR T-cell therapies such as exploiting innovations in synthetic biology and orthogonal receptors to create novel therapeutic signaling networks and utilizing combinatorial treatment approaches to increase response will broaden the impact of CAR T-cell therapy in the next decade. Many of the strategies discussed in this perspective are ready to be tested in clinical trials and have the potential to improve patient responses. Additionally, as the field's mechanistic and molecular understanding of CAR T-cell resistance expands, these new concepts will increase therapeutic possibilities. The combinational use of different strategies described in this article hold promise for superior CAR T-cell therapies designed to expand patients’ options for cancer treatment.

R.M. Young reports other support from Tmunity Therapeutics, and is an inventor on patent applications licensed to Novartis Pharma A.G. and Tmunity Therapeutics, and has received and is entitled to receive future financial benefits from such licenses. C.H. June reports other support from Novartis and Capstan Therapeutics, and personal fees from Poseida, Tmunity Therapeutics, Ziopharm, and Verismo outside the submitted work; is a scientific founder of Tmunity Therapeutics and Capstan Therapeutics; is a member of the scientific advisory boards for AC Immune, Bluesphere Bio, Cabaletta, Carisma, Cartogaphy, Cellares, Cellcarta, Celldex, Danaher, Decheng, Immune-Sensor, Poseida, Verismo, Viracta, WIRB-Copernicus, and Ziopharm; and is an inventor on patents and/or patent applications licensed to Novartis Pharma A.G., Tmunity Therapeutics, and Carisma Therapeutics, and has received and is entitled to receive future financial benefits from such licenses. No disclosures were reported by the other authors.

The authors thank Dr. Neil Sheppard for helpful discussions. C.H. June is supported by NIH grants P01CA214278 and U54CA244711. R.M. Young and C.H. June are supported by a SU2C–Lustgarten Foundation Translational Cancer Research Team Grant and the Parker Institute for Cancer Immunotherapy. U. Uslu is supported by a Mildred-Scheel-Postdoctoral Fellowship of the German Cancer Aid.